4
$\begingroup$

Given two finite dimensional algebra $A$ and $B$ such that $A$ is Gorenstein and $B$ is not. Can the trivial extension algebras of $A$ and $B$ be isomorphic? See http://www.sciencedirect.com/science/article/pii/0022404984900586 1.3. for the definition. Gorenstein means here that the injective dimension of the regular module is finite as a left and as a right module.

$\endgroup$
2
  • $\begingroup$ your link refers to a non-commutative setting, while the Wiki definition of Gorenstein assumes commutative. Are you assuming the algebra to be commutative? or do you use a definition of Gorenstein in a non-commutative setting, for which a definition/reference would be useful? $\endgroup$
    – YCor
    Commented Aug 30, 2017 at 14:11
  • $\begingroup$ @YCor I do not assume commutativity. I added the definition. $\endgroup$
    – Mare
    Commented Aug 30, 2017 at 15:13

2 Answers 2

4
$\begingroup$

If $S$ is a finite dimensional algebra, and $M$ a finite dimensional $S$-bimodule, and if we construct the algebras $A=S\ltimes M$ and $B=S\ltimes DM$, then the trivial extension algebras $T(A)=A\ltimes DA$ and $T(B)=B\ltimes DB$ are isomorphic. This is straightforward to check ($T(A)$ and $T(B)$ are naturally isomorphic as vector spaces, since both are $S\oplus DS\oplus M\oplus DM$, so you just need to check that this isomorphism preserves multiplication), and in fact it is proved in

Wakamatsu, Takayoshi, Note on trivial extensions of Artin algebras, Commun. Algebra 12, 33-41 (1984). ZBL0537.16008.

that if $T(A)\cong T(B)$ then there are $S$ and $M$ as above.

Take $S$ to be a non-Gorenstein algebra, and $M=S$.

Then $A=S\ltimes S\cong S\otimes_kk[x]/(x^2)$ is not Gorenstein, since a finite $A$-injective resolution of $A$ restricts to $S$ to give a finite $S$-injective resolution of $S\oplus S$, and similarly for a finite $A$-projective resolution of $DA$.

But $B=S\ltimes DS$ is symmetric, and therefore Gorenstein.

According to my calculations, the simplest example (taking $S=k[x,y]/(x^2,y^2,xy)$) produces:

$A=k[x,y,z]/(x^2,y^2,z^2,xy)$

and

$B=k[w,x,y,z]/(w^2,x^2,y^2,z^2,wy,wz,xy,xz,wx-yz).$

$\endgroup$
1
  • $\begingroup$ Thanks, I think I need some time to understand that. But I can check now what the trivial extensions of the last A and B are with the computer. $\endgroup$
    – Mare
    Commented Sep 1, 2017 at 11:10
0
$\begingroup$

This answer is just an addition to Jeremy Rickards answer that is too long for a comment. It gives the quiver and relations for the relevant trivial extension algebras using QPA (over the field GF(3)).

Trivial extension algebra of $A$:

Quiver( ["v1"], [["v1","v1","x"],["v1","v1","y"],["v1","v1","z"],["v1","v1","t\ e_a1_1"],["v1","v1","te_a1_2"]] )

[ (Z(3)^0)*x^2, (Z(3)^0)xy, (Z(3)^0)xte_a1_2, (Z(3)^0)yx, (Z(3)^0)*y^2, (Z(3)^0)yte_a1_1, (Z(3))xte_a1_1+(Z(3)^0)yte_a1_2, (Z(3))xz+(Z(3)^0)zx, (Z(3))yz+(Z(3)^0)zy, (Z(3)^0)*z^2, (Z(3))xte_a1_1+(Z(3)^0)*te_a1_1*x, (Z(3)^0)*te_a1_1*y, (Z(3))zte_a1_1+(Z(3)^0)*te_a1_1*z, (Z(3)^0)*te_a1_1^2, (Z(3)^0)*te_a1_1*te_a1_2, (Z(3)^0)*te_a1_2*x, (Z(3))xte_a1_1+(Z(3)^0)*te_a1_2*y, (Z(3))zte_a1_2+(Z(3)^0)*te_a1_2*z, (Z(3)^0)*te_a1_2*te_a1_1, (Z(3)^0)*te_a1_2^2, (Z(3)^0)xz*te_a1_2, (Z(3)^0)yz*te_a1_1, (Z(3))xz*te_a1_1+(Z(3)^0)yz*te_a1_2 ]

Trivial Extension of B:

Quiver( ["v1"], [["v1","v1","w"],["v1","v1","x"],["v1","v1","y"],["v1","v1","z\ "],["v1","v1","te_a1_1"]] )

[ (Z(3)^0)*w^2, (Z(3)^0)wy, (Z(3)^0)wz, (Z(3))wx+(Z(3)^0)xw, (Z(3)^0)*x^2, (Z(3)^0)xy, (Z(3)^0)xz, (Z(3)^0)yw, (Z(3)^0)yx, (Z(3)^0)*y^2, (Z(3))wx+(Z(3)^0)yz, (Z(3)^0)zw, (Z(3)^0)zx, (Z(3))wx+(Z(3)^0)zy, (Z(3)^0)*z^2, (Z(3))wte_a1_1+(Z(3)^0)*te_a1_1*w, (Z(3))xte_a1_1+(Z(3)^0)*te_a1_1*x, (Z(3))yte_a1_1+(Z(3)^0)*te_a1_1*y, (Z(3))zte_a1_1+(Z(3)^0)*te_a1_1*z, (Z(3)^0)*te_a1_1^2 ]

$\endgroup$
3
  • 1
    $\begingroup$ I've checked by hand that those are isomorphic. If you eliminate the relations that say that the arrows all have square zero and commute with one another, you're left in both cases with relators in four generators $p,q,r,s$ of the form $pq,qr,rs, ps, pr-qs$, and for $A$ there are three other degree three relators that are redundant. $\endgroup$ Commented Sep 1, 2017 at 16:32
  • $\begingroup$ By the way, does GAP confirm that $A$ is not Gorenstein but $B$ is? $\endgroup$ Commented Sep 1, 2017 at 16:35
  • $\begingroup$ @JeremyRickard thanks, yes A is not Gorenstein and the other is selfinjective. Thanks for the great answer. Looked like hard work. $\endgroup$
    – Mare
    Commented Sep 1, 2017 at 16:41

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .